Folding Forum

Sadly I don't have a Linux machine to test it right now, but try running a WU with 4 threads, and pin them in the following configuration: thread 0 - cpu 0 thread 1 - cpu 1 thread 2 - cpu 2 thread 3 - cpu 2 You will quickly notice that threads 0 and 1 still keep using 100% of CPU time, although com...

After looking at this again, I think it doesn't matter. RDTSCP timings are used for DLB (Dynamic Load Balancing), which only matters on distributed systems that use multiple MPI ranks. FAH does not currently use multiple MPI ranks, so it probably gets no benefit from better timings. The timings are ...

It does help, thank you. I'm surprised that the bandwidth is so low. I will test my system with different driver versions to see if it might be a driver issue.

The shortage is temporary and there are no plans to explicitly require 4 threads. Even the requirement of 3 threads is simply a temporary workaround due to a bug in the client that only happens on systems with 2 threads.

I decided to try out a 5090 on Windows. I set the power level as low as I could, which was 66%, so it runs around 400W. Seems to consistently get WU's, but sometimes the PPD will say 40-45M, where other times it's more like 20-25M, depending on the WU. I guess that's where Linux could come in handy...

Try restarting the clients directly with "sudo systemctl restart fah-client". Sometimes that is necessary.

Averages are going to be more useful than maximums, because each sample counts the transfers over the last 20ms, so it will be affected by bursts where the transfer uses up the entire 20ms.

That experience is part of what is behind the GPU "whitelist" that also bans a whole bunch of PCI IDs. Some of those very old ones in the list were banned for returning bad results. That makes sense. If I make a PR, I'll have it so that it does not allow overriding an explicitly banned de...

If that's the case then it explains why everything seems like it should work. Maybe the client should give a more meaningful error, something like "GPU detected but not whitelisted" instead of just "supported: false" which could mean anything from not having CUDA drivers installe...

You can save power by turning off the screen even if the lid is not closed.

Don't let the laptop sleep or hibernate while GPU folding or it can hit a bug (actually one bug in the core and one in the client that synergise) that will cause the client to endlessly download and fail WUs.

The DB errors should now be fixed. The GPU WU shortage is being worked on. The CPU WU shortage is being investigated.

Your conventional nVidia drivers won't work with SXM5, I believe. You probably need some sort of pro drivers or something. Not sure how Nvidia separates them Yeah, someone who knows a lot more about it than me I'm sure would have better luck. I sank about 2 hours into trying to get it configured, e...

Thanks, arisu. I'll keep an eye on the forum for news about fixes, but in the meantime I'm pausing my participation due to the performance hit I'm seeing. I should probably replace my machine - it's more than 5 years old - but the budget for elective updates is a bit low. Issue #2 has been fixed an...

The FAH core uses hundreds of times more PCIe bandwidth than it seems like it should and way more than most other CUDA applications (that aren't doing heavy reduction or LLM stuff that is), but even a single PCIe 4 lane (or 2x PCIe 3 lanes) can handle 2 GB/s. I think the reason is just that PCIe is ...

The 5090 uses 575W and can compute about 105 TFLOPS. The 4080S uses 320W and can compute about 52 TFLOPS. So as far as raw compute power goes, two 4080S GPUs is almost exactly as fast as one 5090, but would use more power, at 640W total. That's an extra 65W. If you are on Linux, you can use MPS (htt...